Emit a usable product name from microbench - #1017
Merged
Merged
Conversation
_arch_product_name names the arch spec that microbench generates. Off the two
Instinct memory tiers it recognises, it returned the last word of the device
string, which is not a product name: "AMD Radeon 8060S Graphics" ends in a
marketing suffix, so the generated spec was named "Graphics".
Client parts report the model between the brand and that suffix, so it is read
from there instead. The memory tier is left alone and still covers the generic
strings ROCm containers report -- "AMD Radeon Graphics" at 192 GB has nothing
between brand and suffix, so it does not match and resolves to MI300X as before.
Changes:
- The regex only matches the "AMD Radeon <model> Graphics" form, so discrete
boards ("AMD Radeon RX 7900 XTX") keep the last-word fallback. That form is
what the affected APUs report; widening it to every Radeon string would need
a list of marketing suffixes this does not want to carry.
- Names are joined with "_" so the result is usable as an arch JSON stem, the
way MI300X.json's name field already matches its filename.
- The added test_arch_product_name cases cannot execute anywhere torch is
installed: 6306fc7 dropped the import from _import_microbench, so it returns
an unbound name and the torch-gated half of that file raises NameError. CI has
no torch and skips it, so the suite stays green either way. Restoring those
imports is left to a separate change.
Tested on gfx1151: the generated spec is now named Radeon_8060S.
mgehre-amd
requested review from
ajassani,
devalshahamd,
gabeweisz and
tsrikris
as code owners
September 9, 2026 20:03
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
Contributor
|
@mohbasit for a review |
tsrikris
reviewed
Sep 9, 2026
tsrikris
left a comment
tsrikris
left a comment
Contributor
There was a problem hiding this comment.
Changes LGTM but I'll let Abdul confirm
mohbasit
approved these changes
Sep 10, 2026
tsrikris
enabled auto-merge (squash)
September 10, 2026 19:22
Contributor
|
@devalshahamd can you approve? |
mohbasit
approved these changes
Sep 14, 2026
devalshahamd
approved these changes
Sep 14, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
_arch_product_name(TraceLens/PerfModel/benchmarking/microbench.py) decides thenamefield of the arch JSON thatmicrobenchgenerates. It had two paths:>= 280 -> MI355X,>= 180 -> MI300X).For client parts, the device string
"AMD Radeon 8060S Graphics"is turned into{ "name": "Graphics", ... }This PR fixes it to extract
"Radeon_8060S"